List your company

Task-Oriented Multi-Agent Communication on MMLU, ARC-C, GSM8K, OpenBookQA, MedQA Average (test)

79.03AccuracyTodyComm
49.07856.85464.6372.406Feb 3, 2026

Evaluation results

MethodLinks
TodyCommAttack rate=< 50%2026.0279.03409.2
TodyCommAttack rate== 50%2026.0277.45422.2
Agent PruneAttack rate=< 50%2026.0277.04383
GDesignerAttack rate=< 50%2026.0275.17607.8
TodyCommAttack rate=> 50%2026.0273.39430.2
Complete GraphAttack rate=< 50%2026.0270.92758.8
Random GraphAttack rate=< 50%2026.0268.95606.2
Agent PruneAttack rate== 50%2026.0260.52444
GDesignerAttack rate== 50%2026.0259.05643.4
Agent PruneAttack rate=> 50%2026.0255.78458.2
Random GraphAttack rate== 50%2026.0255.38635.4
Complete GraphAttack rate== 50%2026.0255.19773
GDesignerAttack rate=> 50%2026.0254.07636.2
Random GraphAttack rate=> 50%2026.0250.7625
Complete GraphAttack rate=> 50%2026.0250.23756.6